In what are likely to be our last saltshaker tests of the cycle, today we examine the latest offering from Public Policy Polling (PPP) with a poll that shows the governor’s race at a photo finish 44% to 44% and Quinnipiac (Qpac) showing Crist up by 1 point at 42%-41%
Even though PPP and Qpac both have this race in a dead heat, these polls are actually bad news for Crist.
First PPP: PPP’s offering is bad news in that it has a party balance showing Democrats outperforming Republicans at the polls by 2 percentage points.
Let me be clear about something: Team Crist has done a very good job of reshaping the electorate and, as of this writing, the electorate is about 8 points better for Democrats than it was at the same point in time in 2010. But even with that, it is inconceivable that Democrats will be up by 2 percentage points when all is said and done. The more likely scenario is that Democrats will be down somewhere between 3 and 4 points. (As of book closings from this weekend, Democrats are down 3+ points and based on prior elections and the number of outstanding likely voters, that margin is likely to hold.)
Further, the PPP survey has 23% of the electorate made up of “Independent/Other” respondents and that is also not likely to happen as they currently make up about 18% of the electorate and are projected to be somewhere between 18% and 20% overall.
Why are these two things potentially bad news for Crist?
First, if we adjust party balance to what is a more likely scenario and cut Democrats by a conservative 4-5 points, based on partisan leanings, Scott’s lead grows to at least +3. Second, we need to do likewise for Independents as this survey oversampled them by around 5 points as well. This will cost Charlie at least another point as he is winning Independents by nearly 16 points in the poll. If we recalculate this sample accordingly and make it look more like the actual electorate, this is a poll that shows Rick Scott leading by around 4 points.
Consider also that PPP used a blended sample (with 20% of respondents participating via Internet) which probably explains why their sample is also much – MUCH – younger than the makeup of those who will actually vote. (The PPP poll has only 28% of respondents over the age of 65, while right now those voters make up nearly half of the electorate.)
In the end, the sample has too many Democrats, too many Independents and it is way too young to be reflective of the reality we know today.
Second Quinnipiac: Without going into the details (again) over why we have some methodology concerns with Qpac (see previous saltshaker tests), this poll’s most glaring problems are similar to PPP in that the partisan breakdown is not reflective of what the final tally is likely to be. For purposes of this exercise, let’s take party breakdown from the poll at face value (even though it is unverified and asked in the more unreliable, “Do you consider yourself…?”) and go with the 35% GOP, 35% Dem and 30% Independent/Other.
As noted above, an equal number of Democrats and Republicans voting simply will not happen nor will the electorate be 30% “Independent/Other.”
As such, these numbers need to be taken with a grain of salt as this skewed sample artificially favors Crist. If we were to shift the balance to be more reflective of what is highly likely to happen (add more 3%-4% more GOP respondents and lose 10 points of Ind/Others) and this poll would have had Scott up 2 or 3 points.
Yes, the race is close. Yes, it will come down to the wire. And yes, we are likely to be headed for another Florida squeaker, but c’mon guys, these EAV numbers are available for everyone to see and use. A simple re-weighting of your sample or better tracking of respondents is really not that hard and should have been done prior to releasing either of these flawed numbers.
Verdict(s): For these reasons we suggest taking the PPP and the Qpac polls with several well-earned grains of salt.